Can I decrease the time it takes for EIGRP to converge?

[gjstem]

I have a customer that has a router with two links running in parallel in a primary and backup scenario. They both do receve the same routing info via EIGRP. When both links are up, I can see that the router has the routes via both links in the eigrp topology table. The topology table does list them appropriately as successors and feasible successors. Is there anything that I can do to allow the router to pick up feasible successor routes and install them in the routing table any faster? My customer wants the failover to be transparent to his users. Currently, the failover is knocking off his users even though the eigrp process never has to go active for these routes.

any input is appreciated,

[tmoreo]

There are ways to tweek any routing protocol, however if your network is that simple and you do not have stateful failover you may have other issues. How fast is your failover currently? When you say links, what type of links do you mean? Is it dedicated or demand? Frame, ATM, ISDN?

[gjstem]

Two dedicated links:

1)Primary link is - 5 megabit microwave link

2)Backup link is - p2p T1(hdlc)

Greg

[datnguyen]

You can change the IP hello-interval eigrp and the IP hold-time eigrp. EIGRP uses incremental routing updates. It only sends out update if there is a change in the network topology. The problem with sending only updates is that you don't know when a path is no longer available, you have to depends on the hello timer. To expedite the discover of unavailable routes, you need to decrease the hello interval and hold time.

The default EIGRP hello timer is 5 sec. for high speed links such as FDDI, Ethernet, Token ring. 60-sec for low bandwidth--T1 multipoint circuits, ATM multipoint, SVC, or ISDN BRIs.

The default hold time is 3 times the hello interval, 15 seconds (5X3) or 180 seconds (60X3)

Better yet. You have already paid for the T1 (your backup link), why not use it by load balancing both links using EIGRP.

[ruwhite]

What types of links are they? You'll get faster failover if they are point-to-points.... Also, I see that someone suggested below that you could set the hello interval really low--the lowest you can set it at the moment is 1 second, with a 3 second hold time.

Russ

[gjstem]

I did end up tweaking the hello interval and hold time to 1 and 3. The failover scenario now works beautifully.......

thanks everybody for your input

greg